You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I was playing around with the "Video Frame Classification" feature of label studio to see if I could introduce it into my workflow. Right now the annotation mechanism is decent, but the export file feature doesn't seem to export something that could be converted easily into Apple's CreateML annotation json forma for Action Classification.
CreateML Action Classification format is relatively straight forward (source), but it primarily uses one label per line or dict, with a start and end time-stamp of the video (not frames).
CSV
video, label, start, end
<filename as string>,<label as string>,<time-index>,<time-index>
JSON
[
{
"video" : "filename as string.mov",
"label" : "label as string",
"start" : <time-index>,
"end" : <time-index>
}
]
Time indexes can be either:
An integer of seconds, for example: 0, 3, or 5.
A floating point number of seconds, for example: 0.0, 3.14, or 60.5.
A string of minutes and seconds, for example: 01:03.
A string of minutes, seconds, and fractional seconds, for example: 01:03.14.
A string of hours, minutes, and seconds, for example: 05:01:03.
Hope this is something that could be considered in the feature cycle. Thank you.
The text was updated successfully, but these errors were encountered:
I was playing around with the "Video Frame Classification" feature of label studio to see if I could introduce it into my workflow. Right now the annotation mechanism is decent, but the export file feature doesn't seem to export something that could be converted easily into Apple's CreateML annotation json forma for Action Classification.
CreateML Action Classification format is relatively straight forward (source), but it primarily uses one label per line or dict, with a start and end time-stamp of the video (not frames).
CSV
JSON
Time indexes can be either:
Hope this is something that could be considered in the feature cycle. Thank you.
The text was updated successfully, but these errors were encountered: